My first (failed) business cont.

In my previous blog, we talked about how I came up with my business idea and the problem I attempted to solve. As a quick recap, my AAA was designed to automate customer support of meal prep services in the Philippines.

In this blog, we’ll talk about more failures in bot building and lead generation.

Chatbot Design

I wanted to start simple. My idea was to create a simple customer-facing chatbot for meal prep services deployed to Facebook and/or Instagram.

Now, this chatbot should be custom enough that it should have knowledge of everything about the meal prep service. No need for a human customer support. Shoot a question, the chatbot will answer.

Cool, so how did I build one?

Let's start with the bot builder.

I looked into Botpress and Voiceflow, two of the most impressive no-code chatbot builders out there. Eventually, I ended up with Botpress because they had more platforms I could deploy to (e.g. Facebook, Instagram, Slack, etc.)

Awesome, now let's cover LLM integrations.

There is this coding framework called LangChain that helps you build custom LLM solutions. But then again, I needed code for this, but I don’t really want to. I was just a one-man agency, remember? So I used this no-code tool called Flowise. It’s a drag-and-drop platform for building LLMs, built on top of LangChain.

But wait, what's Flowise for?

Without going into too much detail, I was leveraging Flowise to help me connect our three main components:

• Botpress

• Vector Database

• OpenAI's GPT

The Vector Database is where I planned to store the knowledge base of the client. Anything from meal plan menus, FAQ pages, and support documents. OpenAI's GPT, on the other hand, is the LLM engine for the chatbot itself. Combining everything should give us a human-like chatbot that knows everything about a client's meal prep service.

By the way, here’s the chatbot I created. Feel free to play with it!

Now, let’s talk about pricing for a bit. Here's a quick breakdown:

• Botpress: ~$50 per month per client, assuming 6,000 incoming messages per client

• Flowise: ~$12 per month, assuming we use Digital Ocean's cheapest server

• OpenAI: ~$8 per month, assuming 6,000 incoming and 6,000 outgoing messages per client

• Vector Database: ~$25 per month, assuming we use Supabase

All in all, tooling for the agency should cost me approx. $95 per month. Given a pricing package of Php 10,000 or $178 per month, margins' really ain't that bad.

Lead Generation

I followed the teachings of Alex Hormozi from his book $100M Leads. His main approach for finding your first 5 customers is through warm outreach i.e. ask who you know. Problem was, I didn't know anyone who owned or worked at the industry I was targeting. So, I went straight to his cold outreach playbook.

In hindsight, I should've still reached out to my well-connected friends. Surely, at least one of them knows someone.

Now, my niche was meal prep services in the Philippines. The market is small (red flag), so it wasn't too hard to find leads.

STEP 1: I first listed all the meal prep services in the Philippines I could find. There were 24.

STEP 2: Then, I went to LinkedIn and searched for all relevant founders or senior leaders involved in meal prep services. Unfortunately, for some services, I couldn't find anyone relevant. I didn't have much of a choice for them, so I defaulted to just adding the support email of the service.

STEP 3: For the ones who were on LinkedIn, I used the free-tier version of Apollo.io to gain access to their leads database. I inputted the name of the person, and checked if they had an email or phone number.

After the lead generation process, I ended up with 53 total leads to contact.

It wasn't looking good. Such a small market. But I still went for it. (sunk cost fallacy)

Cold Outreach

I created a new email address: [email protected]. Problem was, this email address wasn't warm enough. Pro tip: warming up an email address avoids having your email be thrown straight to spam.

I used a tool called Instantly.ai, signed up for their free-tier membership, and immediately warmed up my new email address. You can learn more about how they do email warmups on their website. But, in essence, your email address gets thrown into a pool of other addresses, and each of the participants send and receive random emails to/from each other. All automated. Pretty neat, if you ask me.

After my free-tier access expired, my email address was warm enough to send cold emails. Yes, I went through the list and sent an email manually. All personalized.

I still have absolutely no idea if the above was the right email content. Was I supposed to pitch the idea right away? Or perhaps check if they have this problem first?

No clue. All I know was I should expect not to close deals on first contact.

In the end, I only reached out to 21 leads, with 10 follow ups. Shit happened, you know. 🤷

Retrospective

Most of the mistakes I did was already covered by the previous blog. But there were a couple of things I haven't mentioned yet.

First was I spent so much time creating an MVP without validating my idea. Yes, I spent more than a week researching about the perfect tools for my chatbot. I knew from the start the chatbot was feasible, countless have already done it! Researching the perfect tools and optimizing for pricing wasn't necessary at all.

My suggestion: don't spend more than a few hours tinkering and building. But, most importantly, validate the idea first, duh.

Okay, the second was another form of procrastination. I freakin' spent a lot of time (more than a day) thinking about the branding and name for my agency. What was I doing!?

I procrastinated to delay myself from talking to leads. Talking to people was hard, but seeing the business fail by seeing no responses was much harder. See, I was married to my niche, problem, and offer. I didn’t have the “fail fast” mindset drilled into me yet.

So yep, learn from my mistakes. Your takeaway should be to focus on what matters for your business. In my case, it was getting customers. Not branding, not logo making, not even MVP building. Remember, no customers, no business!

Hell yeah, that was such a ride. But, at least this was such a cool experience I can look back to! Stay tuned for more stories like these in the coming years. Peace out. ✌️